Aligning regional innovation capability with territorial space use efficiency (TSUE) is important for regional sustainable development. Using data from 2010 to 2020, this study examines their spatiotemporal patterns and matching-type transitions in the Pearl River Delta urban agglomeration through entropy-weighted TOPSIS, a super-efficiency slacks-based measure (SBM) model incorporating undesirable outputs, and four-quadrant analysis. The results show that: (1) Both innovation capability and TSUE increased overall, but innovation disparities widened while efficiency disparities narrowed. (2) The composition of matching types shifted toward a predominance of high–high and innovation-lagging units. A core–periphery pattern was evident, with high–high units concentrated along the Pearl River Estuary and innovation-lagging units widespread in peripheral areas. (3) Matching types exhibited substantial persistence, alongside prominent transitions from low–low to innovation-lagging and from efficiency-lagging to high–high. High–high and innovation-lagging units showed particularly strong persistence, although some reversals also occurred. These findings support differentiated and locally adapted spatial governance: high–high areas should enhance spatial quality and innovation diffusion; innovation-lagging areas should strengthen innovation capability; efficiency-lagging areas should prioritize site-specific industrial-land regeneration and functional restructuring; and low–low areas should improve infrastructure and industrial conditions while gradually cultivating innovation capacity.